Purpose To determine the clinical risk factors for cytomegalovirus (CMV) retinitis in patients with AIDS. Design A case–control study. Participants The study included 120 patients in whom CMV retinitis had been diagnosed from 1990 through 1999 (cases) and 159 patients without CMV retinitis from the same period (controls). All individuals had AIDS and CD4 counts less than 50 cells/μl at the time of diagnosis of retinitis in the cases or on the corresponding date for the controls. Methods Clinical risk factors were determined by history or physical examination. Confounders controlled for included CD4 count, hospital center, and a series of variables to control for confounding by drug treatment. Statistical analysis was performed by multivariate logistic regression. A systematic model-building strategy was developed from assumption testing to model building to model checking. Main variables measured Presence of visual symptoms, retinal microinfarctions (cotton-wool spots), history of opportunistic infections, and risk factors for human immunodeficiency virus acquisition were determined and compared in both groups. Results The following clinical risk factors were significant predictors of CMV retinitis: flashing lights or floaters (odds ratio [OR], 11.42; 95% confidence interval [CI], 3.43 to 38.01), cotton-wool spots (OR, 2.90; 95% CI, 1.01 to 8.29), number of previous opportunistic infections (OR, 1.81; 95% CI, 1.24 to 2.64), previous nonocular CMV infection (OR, 82.99; 95% CI, 6.86 to 1004.58), previous Mycobacterium infection (OR, 3.41; 95% CI, 0.99 to 11.85), and homosexuality (OR, 2.83; 95% CI, 1.13 to 7.12). Conclusions Based on this study, clinical variables have been identified that elevate the risk of CMV retinitis. These findings may be useful to clinicians and health policy experts in developing rational guidelines for screening, examination frequency, and targeted prophylaxis for CMV retinitis in patients with AIDS.